Earlier this year, the release of DeepSeek\u2014an open-source large language model from China\u2014intensified the\u00a0technological competition between the United States and China.\u00a0In response, one U.S. lawmaker proposed the\u00a0\u201cDecoupling America\u2019s Artificial Intelligence Capabilities from China Act\u201d (S.321)<\/a>, which directly promotes \u201ctechnological decoupling.\u201d The University System of Georgia swiftly placed\u00a0DeepSeek on their prohibited applications list<\/a>\u2014citing data security and cybersecurity risks<\/a>\u2014alongside platforms like TikTok and WeChat. This rapid escalation of restrictions highlights how AI tools can become flashpoints for broader geopolitical and commercial tensions.\u00a0<\/p>\n Technological Decoupling: An Overview\u00a0<\/strong><\/p>\n Historically, whenever a rising power\u2019s industries approached U.S. competitiveness\u2014from the Soviet Union\u2019s space and missile technology to Japan\u2019s semiconductor edge in the 1980s\u2014the U.S. responded with economic sanctions and technological blockades. Today, as China\u2019s high-tech sectors advance, Washington perceives a similar threat, prompting an intensified containment strategy. In 2010, the U.S. National Security Strategy called for\u00a0deeper cooperation with China<\/a>. By 2017, however, China was designated a \u201cstrategic competitor<\/a>,\u201d and in 2022 the U.S. described China as the only competitor \u201cwith both the intent and power to reshape the international order.<\/a>\u201d\u00a0\u00a0<\/p>\n Although these doctrines did not explicitly single out AI, they marked a shift from cooperation to rivalry\u2014laying the groundwork for Washington to more aggressively curb China\u2019s advanced technology capabilities. As AI and semiconductors rose to prominence as engines of economic and military power, they became prime targets in the push to contain China\u2019s technological rise.\u00a0<\/p>\n Nonetheless, the profound interdependencies inherent in global supply chains significantly constrain the feasibility of a full-scale economic or trade decoupling. While the U.S. has tried shifting production away from China to nations like Vietnam and Mexico, fully removing China from global supply chains remains a major hurdle. Take Samsung, for example: although its smartphones and telecom equipment are assembled in Vietnam, critical components\u2014such as electrical capacitors\u2014still come from China,\u00a0effectively keeping China embedded in the supply chain<\/a>. Similarly, reshoring advanced manufacturing to the U.S. faces cost, labor, and policy barriers. Albemarle, the world\u2019s largest lithium producer,\u00a0recently paused its planned U.S. refinery project<\/a>\u00a0due to market uncertainties, highlighting the challenges of replacing China\u2019s well-established industrial capacity.\u00a0<\/p>\n As a result, Washington has focused on restricting access to critical components and R&D collaborations to curtail China\u2019s advancements more directly and rapidly than broader trade or industrial relocations.\u00a0<\/p>\n U.S.-China technological decoupling is a multifaceted process, extending beyond export controls and trade barriers to encompass the many ways the two countries collaborate and depend on each other to chart their technological futures. Although the term itself is broad, its real-world implications become clearer when we break it down into three key dimensions: collaboration, dependence, and divergence.\u00a0\u00a0<\/p>\n 1. Declining Technological Collaboration<\/strong><\/p>\n In recent years, the volume and scope of joint R&D projects between American and Chinese institutions have waned. Workshops, conferences, and professional exchanges once played a crucial role in advancing bilateral innovation, yet they are now fewer and more restricted. Cross-licensing agreements for patented technologies\u2014an indicator of mutual trust in sharing intellectual property\u2014have also dropped, highlighting a growing reluctance to pool resources across national boundaries.<\/p>\n 2.\u00a0Diminishing Technological Dependence<\/strong><\/p>\n Alongside diminishing collaboration, both nations are striving to reduce reliance on each other\u2019s core scientific and industrial capabilities. For instance, fewer Chinese publications cite U.S. research compared to a few years ago, and many Chinese firms have sought local or alternative suppliers for advanced components. This trend underscores a strategic effort to insulate domestic industries from external geopolitical risks and potential supply chain disruptions.\u00a0<\/p>\n 3. Divergence in Technological Trajectories<\/strong><\/p>\n Finally, the two countries are embarking on divergent paths in fields like AI, semiconductors, and emerging technologies. Driven partly by national development goals and partly by restrictions on cross-border technology flows, each side invests in homegrown solutions that may ultimately result in parallel ecosystems. Over time, this divergence could lead to distinct technological standards and approaches\u2014further complicating international collaboration and intensifying the strategic implications of decoupling.\u00a0<\/p>\n Why AI Is the Prime Target\u00a0<\/strong><\/p>\n The decoupling argument considers the development of machine learning applications to be a \u201crace,\u201d and a race implies one winner and all others as losers. This framing may not be valid.\u00a0\u00a0<\/p>\n Those who believe in decoupling with China\u00a0cite\u00a0three main reasons. First, national security:\u00a0AI powers drones, autonomous weapons, and advanced military command systems<\/a>. While there will certainly be military applications of AI,\u00a0their ability to guarantee superiority on the battlefield is not so simple.\u00a0Georgia Tech researchers have noted<\/a>\u00a0that \u201cThe conditions that make AI work in commerce are the conditions that are hardest to meet in a military environment because of its unpredictability.\u201d\u00a0\u00a0<\/p>\n Second, economic growth: according to one consultancy study, AI has the potential to\u00a0nearly double the rate of economic growth<\/a>\u00a0over the next ten years. Some believe that the AI leader will dominate the world economically. Aside from the speculative nature of these growth projections, growth will hinge not on technological leadership\u00a0per se<\/em>, but on the ability of businesses and organizations to make productive use of the applications, as\u00a0Jeffrey Ding\u2019s study of technological diffusion<\/a>\u00a0shows. Furthermore, because the digital ecosystem that supports AI applications is highly globalized,\u00a0one country\u2019s\u00a0leadership in AI does not mean the resulting growth will be confined to that country.\u00a0\u00a0<\/p>\n Third, there is a belief that the leader in the Ai race will establish global technological standards, just as the \u201cWintel\u201d PC architecture and the internet protocols previously guided foundational computing ecosystems. Here again, the parallel may not be valid. LLMs and other machine learning applications work at the application layer and do not require the kind of universal compatibility that computer operating systems and networking require. There may be no AI equivalent of Wintel.<\/p>\n The\u00a0Chip Supply Chain and AI Dominance\u00a0<\/strong><\/p>\n AI\u00a0applications come from a robust digital ecosystem built on four foundational pillars:\u00a0computing devices, networks, data, and software<\/a>. Among these, computing devices\u2014particularly advanced semiconductors\u2014provide the essential processing power behind sophisticated AI models.\u00a0The U.S. has imposed targeted, layered sanctions on China across multiple ecosystem components. It has focused most intensively on the semiconductor supply chain, which U.S. leaders see as a critical bottleneck that can be weaponized. By cutting off China\u2019s access to high-end chips, advanced manufacturing equipment and design software, Washington aims to constrain China\u2019s development of\u00a0AI foundation models<\/a>. This strategy reflects broader geopolitical goals: preserving the U.S.\u2019s technological leadership in AI and preempting China from shaping the emerging global norms, standards, and governance structures surrounding AI technologies. The escalating export restrictions under the\u00a0Commerce Control List (CCL)<\/a>\u00a0vividly illustrate why semiconductors have become the central battlefield in this AI-focused technological decoupling effort. Training foundation models\u2014and, by extension, advanced AI applications\u2014requires cutting-edge chip design and manufacturing capabilities, making semiconductors a natural focal point for targeted export controls from Washington.\u00a0<\/p>\n Initially, U.S. restrictions focused on key technology bottlenecks, including specialized Electronic Design Automation (EDA) software provided by companies like Cadence and Synopsys, which are essential for chip design; Extreme Ultraviolet (EUV) lithography equipment from ASML crucial for advanced chip fabrication; and high-performance GPUs such as NVIDIA\u2019s A100 and H100, essential for AI training workloads. Starting in October 2022, exports of these critical technologies to China were explicitly banned. As Chinese firms began adapting to these restrictions, the U.S. expanded its controls. By October 2023, export regulations encompassed modified GPU variants\u2014including NVIDIA\u2019s A800 and H800\u2014initially designed with reduced capabilities to comply with earlier rules. This move eliminated previous compliance loopholes, underscoring Washington\u2019s commitment to restricting even downgraded technological flows. These measures have tightened even further after 2024. Restrictions on high-bandwidth memory (HBM) and advanced chip packaging will significantly hinder China\u2019s domestic capabilities to produce cutting-edge AI hardware.\u00a0<\/p>\n Collectively, these escalating controls create significant operational uncertainties for technology firms operating within or connected to China\u2019s semiconductor ecosystem. Companies face elevated risks related to supply chain disruptions, heightened complexity in long-term R&D strategies, and increased compliance requirements, making strategic planning increasingly challenging amid intensifying U.S.-China technological rivalry.\u00a0<\/p>\n Decoupling or Trade?\u00a0<\/strong>\u00a0<\/strong><\/p>\n Our research builds on\u00a0Han et al.\u2018s method of measuring decoupling and dependence<\/a>\u00a0We find that the level of U.S.-China \u201cAI decoupling\u201d varies significantly across different subfields. In foundational techniques\u2014such as advanced algorithms or machine learning\u2014China maintains a pronounced reliance on U.S. technology. But in application domains like robotics, we are witnessing a reverse dependency from the U.S. on China\u2019s manufacturing capabilities, supply chain efficiency, and large-scale adoption environments. These interdependencies reflect the\u00a0underlying reality that the American and Chinese AI ecosystems remain deeply intertwined, with each side offering strengths that amplify the other\u2019s progress. This mutual dependence underscores how collaboration\u2014formal or informal\u2014can substantially accelerate AI\u2019s overall progress. Conversely, a more severe decoupling risks undermining both countries\u2019 capacity to innovate, by mandating parallel R&D and duplicative infrastructure.\u00a0<\/p>\n Ultimately, despite the push for AI decoupling, the global tech landscape is deeply interlinked. Complete separation is costly and complex, leaving room for partial cooperation on select standards or foundational research. Yet the strategic uncertainty means that businesses must carefully map and hedge against the shifting fault lines of U.S.-China technological competition.\u00a0<\/p>\n Original publication:<\/strong> Georgia Tech, School of Public Policy, Internet Governance Project (IGP) blog \u2013 \u201cUnpacking US-China \u2018Decoupling\u2019 in AI\u201d<\/em> by Xinger Wei & Peiyu Zhan<\/strong>, 20 April 2025. Original URL: https:\/\/www.internetgovernance.org\/2025\/04\/20\/unpacking-us-china-decoupling-in-ai\/<\/a><\/p>\n This work draws on research undertaken by the MIOIR-based authors as part of an exchange visit to the Carter School of Public Policy at the Georgia Institute of Technology in April 2025.<\/p>\n [\/et_pb_text][et_pb_post_nav _builder_version=”3.0.106″ in_same_term=”off” show_prev=”on” show_next=”on” prev_text=”Previous” next_text=”Next” locked=”on” parent_locked=”off” \/][et_pb_comments _builder_version=”3.0.106″ show_avatar=”on” show_reply=”on” show_count=”on” background_layout=”light” header_font_size_tablet=”51″ header_line_height_tablet=”2″ custom_button=”on” button_icon_placement=”right” button_text_color=”#ffffff” button_use_icon=”off” locked=”on” parent_locked=”off” button_text_size__hover_enabled=”off” button_text_size__hover=”null” button_one_text_size__hover_enabled=”off” button_one_text_size__hover=”null” button_two_text_size__hover_enabled=”off” button_two_text_size__hover=”null” button_text_color__hover_enabled=”off” button_text_color__hover=”null” button_one_text_color__hover_enabled=”off” button_one_text_color__hover=”null” button_two_text_color__hover_enabled=”off” button_two_text_color__hover=”null” button_border_width__hover_enabled=”off” button_border_width__hover=”null” button_one_border_width__hover_enabled=”off” button_one_border_width__hover=”null” button_two_border_width__hover_enabled=”off” button_two_border_width__hover=”null” button_border_color__hover_enabled=”off” button_border_color__hover=”null” button_one_border_color__hover_enabled=”off” button_one_border_color__hover=”null” button_two_border_color__hover_enabled=”off” button_two_border_color__hover=”null” button_border_radius__hover_enabled=”off” button_border_radius__hover=”null” button_one_border_radius__hover_enabled=”off” button_one_border_radius__hover=”null” button_two_border_radius__hover_enabled=”off” button_two_border_radius__hover=”null” button_letter_spacing__hover_enabled=”off” button_letter_spacing__hover=”null” button_one_letter_spacing__hover_enabled=”off” button_one_letter_spacing__hover=”null” button_two_letter_spacing__hover_enabled=”off” button_two_letter_spacing__hover=”null” button_bg_color__hover_enabled=”off” button_bg_color__hover=”null” button_one_bg_color__hover_enabled=”off” button_one_bg_color__hover=”null” button_two_bg_color__hover_enabled=”off” button_two_bg_color__hover=”null” \/][et_pb_text admin_label=”Related posts” _builder_version=”3.0.106″ background_layout=”light” locked=”on” parent_locked=”off”]<\/p>\n [\/et_pb_text][\/et_pb_column_inner][\/et_pb_row_inner][\/et_pb_column][et_pb_column type=”1_3″][et_pb_sidebar _builder_version=”3.0.106″ area=”sidebar-1″ orientation=”left” show_border=”off” background_layout=”light” locked=”on” \/][\/et_pb_column][\/et_pb_section][et_pb_section bb_built=”1″ admin_label=”Connect Title” custom_padding=”0px||0px|” _builder_version=”3.0.90″ background_color=”#1d1d1d” module_class=”footer-section” prev_background_color=”#ffffff” global_module=”380″][et_pb_row admin_label=”row” global_parent=”380″ locked=”off” allow_player_pause=”off” parallax=”off” parallax_method=”on” parallax_1=”off” parallax_method_1=”off” parallax_2=”off” parallax_method_2=”off” parallax_3=”off” parallax_method_3=”off” make_fullwidth=”off” use_custom_width=”off” width_unit=”on” use_custom_gutter=”off” make_equal=”off” module_id=”footer-row” _builder_version=”3.0.89″][et_pb_column type=”1_3″][et_pb_text admin_label=”Contact Us Title” global_parent=”380″ background_layout=”dark” text_text_color=”#ffffff” parent_locked=”off” background_position=”top_left” background_repeat=”repeat” background_size=”initial”]<\/p>\n [\/et_pb_text][et_pb_text admin_label=”Contact Us” global_parent=”380″ text_text_color=”#aaaaaa” parent_locked=”off” background_layout=”light” text_orientation=”left” use_border_color=”off” module_class=”contact-us” background_position=”top_left” background_repeat=”repeat” background_size=”initial” _builder_version=”3.17.6″]<\/p>\n +44 (0) 161 275 0924<\/p>\n [\/et_pb_text][\/et_pb_column][et_pb_column type=”1_3″][et_pb_text admin_label=”Find Us Title” global_parent=”380″ background_layout=”dark” text_text_color=”#ffffff” parent_locked=”off” background_position=”top_left” background_repeat=”repeat” background_size=”initial”]<\/p>\n [\/et_pb_text][et_pb_text admin_label=”Address” global_parent=”380″ background_layout=”dark” text_orientation=”left” use_border_color=”off” module_id=”address” parent_locked=”off” background_position=”top_left” background_repeat=”repeat” background_size=”initial” _builder_version=”3.17.6″]<\/p>\n MIOIR [\/et_pb_text][\/et_pb_column][et_pb_column type=”1_3″][et_pb_text admin_label=”Connect Title” global_parent=”380″ background_layout=”dark” text_text_color=”#ffffff” parent_locked=”off” background_position=”top_left” background_repeat=”repeat” background_size=”initial”]<\/p>\n [\/et_pb_text][et_pb_code admin_label=”Social Media” global_parent=”380″ parent_locked=”off” _builder_version=”3.17.6″]<\/p>\n <\/p>\n <\/ul>\n [\/et_pb_code][\/et_pb_column][\/et_pb_row][\/et_pb_section]<\/p>\n","protected":false},"excerpt":{"rendered":" MIOIR Doctoral Researchers Xinger Wei and Peiyu Zhan explore the complexities of US-China AI decoupling, revealing how geopolitical tensions are reshaping global tech interdependence.<\/p>\n","protected":false},"author":6223,"featured_media":2011,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"on","_et_pb_old_content":"","_et_gb_content_width":"","_exactmetrics_skip_tracking":false,"footnotes":""},"categories":[9,11,13,14],"tags":[],"class_list":["post-2009","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-emerging-technologies","category-innovation-and-technology-management","category-responsible-research-and-innovation","category-science-technology-and-innovation-policy"],"_links":{"self":[{"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/posts\/2009","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/users\/6223"}],"replies":[{"embeddable":true,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/comments?post=2009"}],"version-history":[{"count":10,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/posts\/2009\/revisions"}],"predecessor-version":[{"id":2021,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/posts\/2009\/revisions\/2021"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/media\/2011"}],"wp:attachment":[{"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/media?parent=2009"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/categories?post=2009"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-json\/wp\/v2\/tags?post=2009"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"\"](\"https:\/\/blogs.manchester.ac.uk\/mioir\/wp-content\/uploads\/sites\/409\/2025\/05\/Picture1-PX-Blog.jpg\")
<\/p>\nRelated<\/h1>\n
Related posts:<\/h3>
Discussions on Benefit-Sharing and Data Governance<\/h2><\/div><\/div><\/a>
Unlocking the Future: Exploring Facial Recognition's Role in Retail Evolution<\/h2><\/div><\/div><\/a>
UK's surprising R&D stats pose tricky questions<\/h2><\/div><\/div><\/a><\/div>
Contact Us<\/h4>\n
Find Us<\/h4>\n
Alliance Manchester Business School
The University of Manchester
Booth Street West
Manchester M15 9PB<\/a><\/p>\nConnect With Us<\/h4>\n
<\/p>\n
![\"Twitter](\"\/\/assets.manchester.ac.uk\/phase1\/templates\/images\/icon-twitter-50.png\"){kind=icon}
<\/a><\/li>\n![\"LinkedIn](\"\/\/assets.manchester.ac.uk\/phase1\/templates\/images\/icon-linkedin-34.png\"){kind=icon}
<\/a><\/li>\n